Lusnick/kianda-pay-mobile-money-analytics

Analysis of a simulated mobile money operator in Angola, focused on identifying the root cause of transaction failures and turning it into an actionable business recommendation. *Simulated dataset, built to reflect real mobile money operational patterns.* # Kianda Pay — Mobile Money Analytics Analysis of a simulated mobile money operator in Angola (18,000 transactions), focused on identifying and quantifying the root cause of transaction failures — and turning that into an actionable business recommendation. *Note: this uses a simulated dataset built to reflect real patterns in mobile money operations, not production data from an actual company.* ## The Business Problem Mobile money agents need enough **float** (cash + electronic balance) on hand to complete cash-in and cash-out transactions. When an agent runs out of float, the transaction fails, the customer doesn't get their money, and trust in the platform erodes. This is one of the most common operational challenges in mobile money networks across emerging markets. ## Key Findings - **Overall failure rate: 10.1%** but cash-out transactions specifically fail at **17.5%**, since they're the operation most dependent on agent liquidity. - **58% of cash-out failures** are caused by insufficient agent float not fraud, not customer error, not network issues. - Failure rate **triples** with poor network signal (20.8% vs 6.9% with good signal). - Float shortfall is concentrated in specific provinces (Huambo, Huíla, Uíge) and a small set of high-risk agents not evenly distributed. - Fraud flag rate is 2.4%, and unverified-KYC customers show a fraud alert rate 3x higher than verified customers. ## Recommendation Prioritize agent float replenishment in the highest-risk provinces and the top 10 highest-risk agents identified in the dashboard, rather than distributing working capital evenly, this targets the largest, most fixable driver of transaction failure. ## Tech Stack - **Power BI** data cleaning (Power Query), DAX measures, and interactive dashboard - **Excel** source dataset ## Repository Structure ## Repository Structure - `data/` — raw and cleaned dataset - `powerbi/` — Power BI (.pbix) dashboard file - `screenshots/` — dashboard views (since .pbix ca …